Nice to be here, and congrats on your new little! We met back in 2018ish when you visited Berklee during my senior year. I was EIC of the student newspaper at that time. Looking forward to digging more into these posts as they come, and wanted to suggest your checking out former Berklee President Roger Brown's next act after retiring from the college. Cool music non-profit, Salt Lick Incubator, is helping artists reach their next level through funding and mentorship. https://www.saltlickincubator.org/
Zack, congrats on the Columbia fellowship! Clearly an amazing program, and while you'll be in auspicious company, they're lucky to have you, too.
Thanks for the opportunity to weigh in on future Zogblog topics. Here's mine; algorithms.
Full disclosure; this was the subject of both my library science Master's thesis (music recommendation algos actually narrow the suggestion pool), and in a way, my musicology one (sample-based hip hop DJs = expert music recommenders). So I'll try to stay out of the weeds, but my interest is not solely academic.
In fact, it's super cynical; why don't music recommendation algorithms serve the labels better? In other words, why are alogorithms designed to benefit listeners more than music labels?
Bit out of my depth here (and very much in yours), but here's how I understand the basic arrangement: the streaming services pay licensing fees to the record labels. Those platforms charge subscription fees, and offer subscribers music on any connected device.
Of course, among other stuff, streaming services also offer music recommendations.
Consumers like it when recommendation engines come up with suggestions that reflect their tastes; it's a cool perk to, say Spotify (which I use). It's one of several in-platform features; playlists, sharing, exclusive artist content, links to merch, concert listings, etc.
Recommendation is an outlier because it's generative; other features get scraped from web sources (merch, concert dates) or built into the platform (playlists, social links). But unlike those others, recommendation suggests independent knowledge of the subscriber. When it works well, it's gratifyingly uncanny -- it implies a relationship between the service and the listener. For me, part of the appeal is that Spotify seems to be treating me not as just a subscriber, but as a listener -- and to be listening back.
But felicitous recommendations don't drive revenue to Spotify's licensing partners -- record labels. And without a critical mass of content from those labels, Spotify has spotty offerings and disappointed listeners. Surely, no recomendation algorithm will make up for weak selection. Consumers will go for the service that offers every major label's entire catalog, over one that can recommend something interesting that they've never heard before, and may or may not like.
So, if Spotify keeps sending me deep catalog stuff like say, Lee Dorsey, that keeps me happy. But Dorsey, sadly, isn't dropping new records, touring, licensing his music to ad agencies, or otherwise building a productive brand that will make his record company serious money (probably fumbling the lingo here, but you know what I mean). He's a legacy artist, unlikely to enjoy a revival that his label can cash in on.
Labels want consumers directed toward their active, wide-market artists (I assume), not niche or inert ones. For the kid who likes the Time, direct her to Bruno Mars, (who's active on Insta, Twitter, etc.) not dormant, one-and-done Ready for the World.
Last thing -- who would notice? Let's say Spotify (or Apple Music, Tidal, etc) changed their algo, to favor the labels. Now listeners are recommended current artists more often than catalog fixtures. Wouldn't that be the kind of invisible shift that listeners would scarcely notice, while satisfying Spotify's label partners?
The basic conclusion I came to in my own research was that, while listeners like you and me would love it if music recommendation algorithms directed us to unforeseen wonders, we're the only sort of people who care about that. Devoted listeners will always dive deep to seek out interesting new stuff, whether their streaming service uses an enlightened (or agnostic, or unbiased, whatever) algorithm or not. We'll harken to sages from the earlier eras to guide us -- radio DJs, record store clerks, older sisters, sample-based hip hop DJs -- in our sonic quests.
More casual listeners won't really care. If you like Lady Gaga, you should probably check out Mae West, but if Spotify directs you to Lizzo instead, who's gonna be mad? Probably not that Lady Gaga fan.
Anyway, that's my two cents. Love the blog. Keep on doing your good thing.
I offer two apologies; being incapable of writing a short email. And being a stingy non-subscriber. My life has been in a bit of flux since my wife and I moved to Vermont -- she's taken a job in the President's office at Bennington College. Just bought our first lawnmower, for real.
Hey Zack,
Nice to be here, and congrats on your new little! We met back in 2018ish when you visited Berklee during my senior year. I was EIC of the student newspaper at that time. Looking forward to digging more into these posts as they come, and wanted to suggest your checking out former Berklee President Roger Brown's next act after retiring from the college. Cool music non-profit, Salt Lick Incubator, is helping artists reach their next level through funding and mentorship. https://www.saltlickincubator.org/
All the best,
Dom
Zack, congrats on the Columbia fellowship! Clearly an amazing program, and while you'll be in auspicious company, they're lucky to have you, too.
Thanks for the opportunity to weigh in on future Zogblog topics. Here's mine; algorithms.
Full disclosure; this was the subject of both my library science Master's thesis (music recommendation algos actually narrow the suggestion pool), and in a way, my musicology one (sample-based hip hop DJs = expert music recommenders). So I'll try to stay out of the weeds, but my interest is not solely academic.
In fact, it's super cynical; why don't music recommendation algorithms serve the labels better? In other words, why are alogorithms designed to benefit listeners more than music labels?
Bit out of my depth here (and very much in yours), but here's how I understand the basic arrangement: the streaming services pay licensing fees to the record labels. Those platforms charge subscription fees, and offer subscribers music on any connected device.
Of course, among other stuff, streaming services also offer music recommendations.
Consumers like it when recommendation engines come up with suggestions that reflect their tastes; it's a cool perk to, say Spotify (which I use). It's one of several in-platform features; playlists, sharing, exclusive artist content, links to merch, concert listings, etc.
Recommendation is an outlier because it's generative; other features get scraped from web sources (merch, concert dates) or built into the platform (playlists, social links). But unlike those others, recommendation suggests independent knowledge of the subscriber. When it works well, it's gratifyingly uncanny -- it implies a relationship between the service and the listener. For me, part of the appeal is that Spotify seems to be treating me not as just a subscriber, but as a listener -- and to be listening back.
But felicitous recommendations don't drive revenue to Spotify's licensing partners -- record labels. And without a critical mass of content from those labels, Spotify has spotty offerings and disappointed listeners. Surely, no recomendation algorithm will make up for weak selection. Consumers will go for the service that offers every major label's entire catalog, over one that can recommend something interesting that they've never heard before, and may or may not like.
So, if Spotify keeps sending me deep catalog stuff like say, Lee Dorsey, that keeps me happy. But Dorsey, sadly, isn't dropping new records, touring, licensing his music to ad agencies, or otherwise building a productive brand that will make his record company serious money (probably fumbling the lingo here, but you know what I mean). He's a legacy artist, unlikely to enjoy a revival that his label can cash in on.
Labels want consumers directed toward their active, wide-market artists (I assume), not niche or inert ones. For the kid who likes the Time, direct her to Bruno Mars, (who's active on Insta, Twitter, etc.) not dormant, one-and-done Ready for the World.
Last thing -- who would notice? Let's say Spotify (or Apple Music, Tidal, etc) changed their algo, to favor the labels. Now listeners are recommended current artists more often than catalog fixtures. Wouldn't that be the kind of invisible shift that listeners would scarcely notice, while satisfying Spotify's label partners?
The basic conclusion I came to in my own research was that, while listeners like you and me would love it if music recommendation algorithms directed us to unforeseen wonders, we're the only sort of people who care about that. Devoted listeners will always dive deep to seek out interesting new stuff, whether their streaming service uses an enlightened (or agnostic, or unbiased, whatever) algorithm or not. We'll harken to sages from the earlier eras to guide us -- radio DJs, record store clerks, older sisters, sample-based hip hop DJs -- in our sonic quests.
More casual listeners won't really care. If you like Lady Gaga, you should probably check out Mae West, but if Spotify directs you to Lizzo instead, who's gonna be mad? Probably not that Lady Gaga fan.
Anyway, that's my two cents. Love the blog. Keep on doing your good thing.
I offer two apologies; being incapable of writing a short email. And being a stingy non-subscriber. My life has been in a bit of flux since my wife and I moved to Vermont -- she's taken a job in the President's office at Bennington College. Just bought our first lawnmower, for real.
Cheers,
damien
Congrats on your daughters birth.